Articles by Melvin Manoukian — Page 23

Claude Sonnet 5 launches as Anthropic's new default model at 2 dollars input, cheaper than GPT-5.5 and Gemini 3.1 Pro on the agent segment.

DeepSeek releases DSpark, a speculative decoding framework boosting model speed 60 to 85% on GPU. An open-source statement.

Meta bans engineers from using Claude Code and Codex outputs, fearing rival AI capabilities will leak into its own training data.

Google opens Gemini Nano Banana personalized image generation to all US users for free. Europe stays locked out.

California gets Claude at half price for state agencies. Newsom becomes the first US governor to institutionalize Anthropic in government.

Ford rehires 350 senior engineers after its AI quality systems disappointed, saving hundreds of millions on warranties and avoided recalls.

Qihoo 360 launches Tu Long Feng, its bug-hunting AI, and its CEO frames Claude Mythos as a cyber-nuclear weapon of the AI era.

Dawn Song joins Meta as VP AI Research and says AI agents that deliver real economic value are the industry's next milestone.

SK Hynix and Samsung commit $590 billion to AI memory as global prices are expected to climb 40 to 50 percent through 2027.

The Tencent Study from Youtu Lab shows GPT-4 only completes 14 percent of WebArena tasks, and agents miss more than half of their own skills.